In V3, as per your documentation , you have implemented a method to create Q&A pairs from our data for RAG.

I have a question about this. If a document generates say, 100 Q&A pairs and a user asks a complex question whose answer is spread across say, questions 1, 23, and 87, will the RAG technology retrieve all these three questions and combine their answers to provide a complete response? Or will it only provide a partial answer based on the single Q&A pair it selects?